nagy%warner.hepnet@LBL.ARPA.UUCP (06/08/87)
>Geoff Hill UUCP: watmath!thunder!gihill >Computer Services BITNET: gihill@lakehead >Lakehead University PHONE: 807-345-2121 (400) >Thunder Bay - Ontario - Canada >1. Can the SYSDUMP.DMP file for a satellite node be moved onto the > satellites local disk or must it stay on the boot nodes's common > cluster disk ? I know you can do this for the satellite's PAGEFILE.SYS > and SWAPFILE.SYS. SYSDUMP.DMP must be on the boot disk, if you want SYSDUMP.DMP. If there is no SYSDUMP.DMP file, no dump of physical memory will be saved on a system crash. On the other hand, you can do what was done on the FNAL VAXCluster here (large CI cluster with 8800, 2x8650, etc.); to reduce usage of common system disk, a single copy of SYSDUMP.DMP was placed in the SYSCOMMON root. This file is large enough to contain the physical memory of the largest system (one 8650 has 48 MB). After any system crashes, SYSDUMP.DMP contains the memory image of that system (this is "rescued" using ANALYZE/CRASH if it needs to be saved). If multiple systems crash simultaneously then the file may contain garbage. So far, it has worked very well and released about 108 MB of disk space used by SYSDUMP.DMP files in the various roots. >2. We have a TU81 tape drive attached to one of our satellites. > Is there any way to make the other cluster nodes recognize the > tape device (like a cluster-wide disk) ? The is no way to make a TU81 available cluster-wide (cluster-wide tapes are only available via the HSCs on CI clusters). Digital mumbled about the possibility of a tape MSCP server appearing someday for the VAXClusters (this was at a session 2-3 DECUS's ago). >3. I have heard rumours that VMS 4.6 will allow you to set up > a satellite node's local disk as an alternate boot node disk. > Can anyone confirm this ? Can be done now, is just "not supported". The trick is coordinating things to avoid deadlock conditions (i.e., system A is booting and requires system B's disk; system B is booting and requires system A's disk) and what to do if only 1/2 of your boot nodes are up... I have heard the same rumors that with V4.6 the VAXCluster node limits will be relaxed and that LAVCs will support multiple boot members. = Frank J. Nagy = Fermilab Research Division EED/Controls = HEPNET: WARNER::NAGY (43198::NAGY) or FNAL::NAGY (43009::NAGY) = BitNet: NAGY@FNAL = USnail: Fermilab POB 500 MS/220 Batavia, IL 60510